Improving existing processes is what innovation is all about. When the eldest son of Tripler Jeroen returned from his driving lessons with yet another piece of paper filled with several different drawn traffic situations, it got him thinking. A combination of the physical world and IT should make driving lessons more interactive and effective. By recording a video at important moments during the driving lesson, the student can learn from these situations and efficiently direct his/her driving training.

We managed to find a way to make this efficient learning method practicable. By placing a 360-degree camera on the roof of the training car, the tool continuously streams images to a tablet with a dedicated   native Android app in the car. If a learning moment comes up during the lesson, the driving coach can describe it, but viewing the images is much more efficient for the student.

If a complex roundabout comes up on the route or a pedestrian suddenly appears out of nowhere while the student is driving, the instructor can activate the i-Cam with one push of the button. The images of the last 20 seconds are recorded and together with the 20 seconds after that, they form a visual clip. In this video, the student sees the trigger, the context and his or her actions, clarifying immediately what could be improved.

This way, a driving coach can offer students the opportunity to better prepare for the next lesson or exam, so they can be more relaxed in the car. By incorporating feedback into the recorded clips, and then uploading them to a personal web environment where privacy is guaranteed, will benefit the student, the coach and the driving school

Being able to place a well-functioning 360-degree camera on top of the training car, required a substantial amount of creativity. We built our own native Android app as a dashcam and integrated it with an existing 360-degree camera. Both the students and the driving schools have a separate mining environment.

When realising the camera on the roof of the training car, we ran into heat problems. After some trial and error, we came to the conclusion that we needed a new roof plate. This piece of tailor-made Triple artwork has a built-in cooling system and a 3D printed holder for the camera.

Blurring faces and license plates is not a new technique, but applying it to 360-degree videos requires technical creativity and ingenuity. Existing sensor software did not work optimally because the images within a 360 sphere were so distorted that it did not meet our requirements. In the end, we used our own image detection model and trained it with our own 360-degree video data.

In addition, we used our own Container Orchestrator with Azure Container Services. The code of the i-Cam blur 'engine' is therefore only switched on when there are uploads and the back-end makes considerably more efficient use of resources.

The i-Cam is now being tested extensively by the ANWB.